Transcript of the talk, from a Whisper transcription of the audio. Auto-generated: speaker names in particular are unreliable. ======================================================================== # Precautionary Sourcing and the Macroeconomics of De-Risking Authors: Ester Faia Discussant: Doireann Fitzgerald Video: https://www.youtube.com/watch?v=wEt4iKiNNZQ&t=16381s ## Talk (04:33:01 – 05:17:00) [04:33:07] With the first session of the afternoon, so where do we have a step? [04:33:32] I was going to talk about precautionary sourcing and the macroeconomics of. [04:33:38] So. [04:34:04] Actually what George was saying before, then when we go across. [04:34:11] When we go across sectors and across origins, we see like different type of movements in some areas in some regions, we see retrenchments and in others we see like going back. [04:34:22] So for instance, if we, I mean this takes census data and contract data and divides the goods, hs for by whether they were tariffed according to the 301 or not. [04:34:36] And the line is around 2018. So the level of the tariff and you see that it goes down, but actually goes down the yellow, which is the ones that are not tithed. [04:34:46] And the red ones, they don't, they don't go down. And then if at the same time, this just suggesting evidence is not yet causal identification, but I will go to causal identification, Bartek and so on and so forth. [04:34:57] So if at the same time you plot the caldara uncertainty and tariff uncertainty index, and I use the monetary policy shock by a buck and room and dailies as a placebo, you see that this thing actually went up before the tariff level. [04:35:14] And there is no other trend in monetary policies. So what this already suggests to is that actually much of the retrenchment, at least for China, but we will see also for the allies. [04:35:24] It's not so much due to the level, but it's due to the increase in uncertainty. And I mean this will be confirmed once I go to the casual evidence. [04:35:35] So, yeah. [04:35:39] I'll, it's, it's, I ought to go to lies the cause that a shock. Let me get there. But yes, I used, you know, most refined way to get an exogenous shock. [04:35:51] And tariff uncertainty, the caldara is a tariff answer. Yes, of course, of course, you can. No, no, no, no, it goes up and down. [04:35:59] It's like, but you need to autogonalize it to have causal identification. Yeah. [04:36:03] Okay, so then I mean motivated by this, what I started to think is, you know, how, how you use I mean I started really from the network. [04:36:13] And I mean, we know we already have the bakwai and fari, of course, which is an exogenous network and the goal of that paper was to study misalocation in the long run. [04:36:22] But one way to think about how we measure tariff policy or the impact of tariff on on on the network on the global value chain is to take into account the Lucas critique. [04:36:31] So if you take the, if you keep the network fixed, the model will be subject to the Lucas critique because effectively people are moving out. Okay. [04:36:40] Now there is a paper actually that applies this more to trade, which is a medical paper by Coppitoff and other, another group of trade quarters that studies endogenous network so it basically endogenizes the shares that in the bakwai and fari are exogenous but on the [04:36:59] market. Now let me say this paper, my paper is not say that this is wrong. They had just a different question. Their question was long run misalocation. They have a terogeneous misalocation. [04:37:10] So that's perfectly fine to keep it fixed. Okay. But now the question is, if I want to see what is the impact on the tariff and I see that in response to shocks firms are reacting then I have to see exactly how much they are reacting. [04:37:23] So I have to see where I come in and basically I endogenize the bakwai and fari but under assuming an insurable risk. Let me also clarify what is very important that I'm not thinking about macro uncertainty but I'm really thinking about firm level uncertainty. [04:37:38] So this is Julia Iagari shocks at firms level why because otherwise I wouldn't have the heterogeneity. Okay. So in the network of course there is heterogeneity across sectors in terms of production. [04:37:48] There is no heterogeneity and there is no precautionary channel. If you don't put the uninsurable risk. So, of course, the first temptation is just to put an uncertainty shock at macro level but this is not what I want to do. [04:37:59] And what it will lead I mean it's a theoretical, theoretical contribution. I will derive risk misalocation wedge. So, and parallel to bakwai and fari that have a production wedge I will derive a distribution of this risk misalocation wedges which [04:38:17] exists also in the half model I have a paper where we show this in the half. [04:38:29] What is the paper and pencil. Okay, so it took me quite some times but but but then afterwards I decided to bring it to more quantitative framework so I put it into a macro model with heterogeneous agents so the agents are also subject to the same risk as the firm so it's a full [04:38:53] hunk model. Not you can there's no sticky prices there is no bells and whistles. There is just a network and and and and you will see a bunch of other things that are needed there. [04:39:04] But in order to discipline this model I start by having by pulling some evidence and the evidence I'm looking for casual identification because the model will match exactly the same elasticity. [04:39:14] So I start with sector level evidence from census and com trade. Then I use this Punjiba data which is B2B relationship across the world and I use only the US sample where I do an event study analysis and then an anticipation design, which I created by importing into a local projection a [04:39:33] bar take type of index bar take type of instrument that we use with Shabnam and securities but here I do it with that with the shares. [04:39:42] Okay, so then then I go to the theory so I will show you how I derived in the generous network the firm will have to. [04:39:49] But anyway, what one thing I forgot to say so the result that I get there is that the tariff level doesn't do anything. The answer the anticipation design tells you that tariff uncertainty shocks drive everything but geopolitical shocks because I also put that Georgia. [04:40:05] So don't don't do anything other. Okay, so it's surprisingly but once I correct for pre trends they don't do anything. Okay, and then when I go to the theory I first do the paper and pencil where I show that even if you have an equal markup so like in the [04:40:21] previous slide, you can see there are heterogeneous production wedges so even if you have an equal wedges if you if firms have purely purely type of shocks they have this precautionary saving and this the correlation that pops up creates this misallocation of misallocation of production. [04:40:39] So yeah, the whole idea that will go throughout the paper and it's very very simple is the following. So in trade, we are used to think that firms choose a location or a sector I mean or an import, just because based on on the efficiency so it's low cost or its production [04:40:56] here the intuition is going to change completely. So the firm will trade off the production complementarity so what is efficient with the risk. Okay, so that's that's the type of trade off that will come in. And so what will happen is that when risk goes up in one origin and in one sector, the firm is willing to move let's say from [04:41:14] China to Mexico, although Mexico is less efficient. And that is where the misallocation loss comes. Okay, and the misallocation loss is of course a distribution across all firms according to their capacity to had risk. [04:41:26] Okay, so then I go to the quantitative model that where I have eight sectors and I mean I put three origins I will be very specific about this, but the whole point of the quantitative model is to first of all, endogenize the kernel because much of what drives the [04:41:41] precautionary savings of the firms in the partial equilibrium model is their kernel so how they wait risk. And in the generic in the models, there will be workers in one way that I linked the workers to the firms is through a limited commitment contract a lot of the [04:41:57] money that is not available to the firm. So that basically transfers the risk of the firms on the workers and so the workers becomes themselves. [04:42:06] And it originally is agents, and I calibrate this on the with a pistol ferry and skiver the evidence. Okay, I will skip the literature review it touches many things. I would say the only macro paper that is you know looking at also [04:42:20] is a job name and of course she has the network, but just focusing much more on monetary policy I'm not going to teach any monetary policy no sticky prices no sticky which is not I mean it's just streamlined macro model. [04:42:34] Okay, so, in the in the evidence I will briefly tell you two things and then go directly to the anticipation channel. So as I said first I do a very simple, you know, different deep with the with the sensors and the [04:42:47] on trade so I regress the product level share from China on on on whether the on on post tariff so post the 2018 and whether the product was started for not sequentially after 2018 and what I see there is basically there is a big retrenchment from China [04:43:06] particularly electronics and machinery is and there is an insuring into Mexico and Vietnam but also firms are bringing production back. Okay, this this will come out a lot. [04:43:19] And when I use the Punjeeva level data I do a classical event study analysis. Okay, except that is weighted by exposure. So here I have the firm level a share of importers, either from China or from the allies, let's say Mexico, Vietnam, and the others. [04:43:33] And here there is I mean this is the event study so when the firm is hit when it's production when it's intermediate inputs are hit by the tariff and this is proportional to the exposure. Okay, I could have like 80% or 30% that that I'm importing now. [04:43:50] When I first started to see the event studies I was very enthusiastic because it goes boo down with China and it goes boo up with Mexico and Vietnam. But then as usual with the event study do start to correct for pre trends and when he I put the pre trends. [04:44:06] Everything disappeared. Okay, and so now luckily my Berkeley voters taught me to read negative results as conducive to positive results. And so here is the positive results meaning that then you know when you have pre trends everything and men's that is driven by anticipation and so what I do here is I device is anticipation design leveraging on the caldera shops. [04:44:33] It's a classical local projection. [04:44:35] So the change in the share is still Punjeeva data the change in the shares of each importer from each origin. Again, let me repeat this is firm level evidence. Okay, so it's it's really granular. I mean, by the way, there are so many observations that I can only run it with the plugin in WRDS. [04:44:52] It doesn't go on my computer. So this is from from each origin. I mean there is, of course, firms and origin sector and origin fix effects that is time fix effects that is absorbing the level of the tariff. And then this is the Bartik design. [04:45:07] Okay, so I extract I ought to go and allies the caldera tariff uncertainty shock and the caldera geopolitical shock with the same methodology. No, I don't know. I mean I, I thought somebody had to go and allies it but nobody did it. [04:45:21] And I use exactly the same methodology that we use in monetary policy of fiscal policies so the power and swans and basically. And this is the shock and it's weighted by the exposure so how much the firm is exposed to that origin and the exposure is kept fixed to the average pre pre tariff. [04:45:38] So it's basically it's a Bartik shift shift share with the goldsmith pink come type of a cause of genetic what doesn't mean so typically an instrument has a shock and then exposure. Okay, again, I have a paper which I'm name on security is doing this. [04:45:52] Now the body you have to keep one of the two things fixed to have the identification and the other one has to move. Okay, in the goldsmith goldsmith pink and word and the exposure is fixed and the shock moves. [04:46:05] Okay, in the board with the cool and gravel is the opposite. I'm using this. Okay, so now the results. So here you see basically the response of the shares from China. This is only from China, not from the other countries in response to a trade policy uncertainty geopolitical [04:46:20] uncertainty and this is the placebo so I put one in a policy to see whether this is driving things you see the placebo flattened down so it's a placebo and this one goes significantly down. [04:46:32] And let me also tell you I mean I have it on the table if you don't believe it but the typical thing to check that there are no pre trends is that you look at the coefficients before the event and they're not significant and in fact they are okay but the geopolitical [04:46:44] uncertainty I want to just say like before I was your token allizing it. Yes, just give me a second before to go in a lighting it looked significant but then I were to analyze and it became nonsense. Okay. [04:47:24] The king. I mean I kind of know you and so if so this one is the event study. Sorry. [04:47:40] Okay. [04:47:43] And study, which is what I did before you see the king because it's around one level. Okay, but when you have a local projection, you're responding to a sequence of shocks. So technically is the same thing of an event study is just that it's a sequence of [04:48:07] shocks, but it's the same logic for identification so it's a party. [04:49:06] And so I can have it only for China or I can pull it across different countries. Okay, and here is pulled. And so you see that the impact is actually much large. [04:49:17] Okay, so this is an elasticity. It's casually identified and it gives me a number. Okay, which I can then I mean I can have my model my quantitative model then. [04:49:28] Okay, now, well, I mean here there are a bunch of restrictions of course and how you do and how you select a sample and I try different things are very, very long appendices on this. [04:49:39] Okay. [04:53:01] Right after that. Yes, you will see it in a second because I derive the wedge. Okay, so now let me finish first the two maximization problem because now you will see the optimization of the alpha and you will see how alpha depends on the wedge. [04:53:13] Okay, that will answer your question. Okay, now, first of all briefly, it chooses the first stage chooses like in copy to the shares to minimize the expected cost where this is the kernel and this firm specific because there is an insurable risk. [04:53:28] Okay, so the cost of the unit cost and the quantity. It turns out that I mean I can use Jensen's inequality in that log algebra so by Jensen's inequality, I take the expected value and I can write the maximization problem of that expected cost as the complementarity. [04:53:46] Okay, minus the sum of the entropy basically the risk weighted expected cost of each input. Okay, so this sum has the price of each input plus some covariance here as the risk misaligned wedge. [04:54:03] Okay, if there is no first stage like in black wine party of if there is like perfect market every complete markets every firm has the same kernel so these things drops out or at least is the same for every firms but in the same plan of idea that is something that differs in the choice of every firm and here the choice of every firm. [04:54:24] I mean here the firm is balancing what I was telling you before the trade off between a certain alpha I is increasing my complementarity my productivity but at the same time is increasing my expected cost. [04:54:37] Now the expected price is the same for all firms that are buying a certain input, but how that input affects me depends only on me. [04:54:46] So firms that are less they have less ability to hedge when the China shock comes, they are hit more than others. Okay, so this is really firm specific that's where the misalocation comes from. [04:54:56] But there is no these differences in my cops or in anything else like in the black wine party. Now I can of course, you know also computed exactly and link it to the to the. [04:55:07] Yeah. [04:55:09] Yeah. [04:55:29] Yes, yes. [04:55:35] I can assume that is heterogeneous like in black wine party that was an assumption of course right because we know that even when we estimate markups they are a sector but the point of this paper and pencil is just to show you why risk creates a miss a production [04:55:50] misalocation. Okay. [04:55:51] Yes. [04:55:54] I have the market but is it homogenous. Okay. [04:55:59] And then just catch up so is that you can of course also measure should name the wedge by taking basically the weighted the risk price minus the risk price under the perfect complete market which is the copy tough and it comes out that is exactly equal to the covariance so it makes very much sense. [04:57:11] If I am complimentary to China, or if I can hedge particularly well, then this correlation might be even negative in fact in the helpman papers. [04:57:20] Most of the diversification from trade comes from the fact that some of those correlations are not although that doesn't come from an insurable. [04:57:28] Yeah, yeah. [04:57:30] Sure. [04:57:33] I was pretty obvious because when you take the first recognition, you see that from a certain input, there is some productivity gain. [04:58:37] So if I, if I am an Italian textile and I want to import garments from Bangladesh, special garments from Bangladesh, this thing is going up. [04:58:48] But then if it's not in comes to Bangladesh, this expected risk goes up. [04:58:53] The key thing is that this expected risk is the expected price which is the same for everybody because there is no aggregate uncertainty and this part that loads on only on those producers. [04:59:05] A quiet and funny. And let me go here so maybe I can. [04:59:18] So in the back, why. [04:59:21] Some of the. [04:59:45] The net is coming from the production function, not from risk. [04:59:50] Well here, on top of the heterogeneity of the production function, there is also the base. [04:59:55] Okay, okay. [04:59:56] Perfect. [04:59:57] All right, so now let me go a little bit quickly because I think. [05:00:02] Okay. [05:00:03] Yeah, I just want to briefly say because this is considered quite important in the misallocation literature. So one job this distribution of wedges, you want to know how much this impact the aggregate. [05:00:14] Okay, that's the whole point of the she and clean or that misallocation is important for the aggregate. [05:00:18] And so what you can do is that you can compute the share per sector, the difference between the one in the into the imperfect insurer economy with respect to the one with complete market. [05:00:30] And this is the towel the wedge that I told you are firm level and in aggregates up basically with this with this with this second derivative. [05:00:39] And, and then you can also compute what bakwa and fari compute the donor weighted. Okay, that tells you how imagine again that a firm is hit by risk. [05:00:48] Then all its input that are given to other producers are carrying that risk and it goes over. Okay. And those are basically carried by the donor weights. [05:00:58] And here you can have the donor weights and the day in perfect economy and the donor weights on the perfect economy, which I derived by taking a planner object quadratic objective function. [05:01:09] This is also like in copies of so I'm not inventing anything new. And what comes out is when you when you aggregate all those wedges up and all those weights up and you take the difference at the aggregate level. [05:01:23] Again, it's going to depend on this metrics T, which is a network corrected Houlton loss so it's weighted by this by this donor weights back inside it contains the wedges of every single firm. [05:01:38] Okay, it's just that those wedges keep multiplying with the donor weights. [05:01:45] I did not have capital here I'm adding exchange rate risk and capital. Okay, so I'm assuming that some of the origins are foreigners and so on top of the productivity risk per sector. [05:02:30] There is also an heterogeneous and exchange rate risk and I can also add some tariff risk. Okay. And the other thing that I do when I go to the larger model, I start to distinguish between shares per sector and per origin. [05:02:43] Okay, so I can import metals or machineries and I can import them either from China or from Mexico from the top. Okay, that pen paper and pencil model was not distinguishing those two things. [05:02:54] Because I didn't know how to solve it. [05:02:56] And. [05:02:57] Okay, so then I mean I'm using something that in trade is actually very, very old. There is an old book by Anderson and the Palmer that a quarter of mine for material mine, you're my good to be animalist to me. [05:03:11] And that shows that there is an equivalence between this entropy problem that I showed you before where you minimize the entropy like in robust control and a logistic and a logit optimization with the gamble shocks. [05:03:25] Okay, there is an equivalence between the two. It's actually shown also in the monopsony literature by her keynote and monkey. I mean, it's very well known results. [05:03:34] Okay, so when you go into the more quantitative model, you can use this type of maximization where this is the share now I'm changing notation is not an alpha it's an S and the S includes the sector shares and the origin shares. [05:03:49] Okay, and those are the matching parameters or the productivity complementarity. And this is the risk as before. Okay, so it's the same thing. [05:03:58] So in order to get the share of each firms investing in each origin and in each sector. I'm using the gamble shocks. The gamble shocks that tends to smooth this me clarify what is the difference with before before in the paper and pencil. [05:04:13] When there was a zero, I didn't care. Okay, but here when I go into the numerics when you know before an alpha could be zero. That's fine because it's paper and pencil. [05:04:23] When I put it in the code, the code jumps. Okay, and so I need to smooth it and I smooth it with the gamble shocks, leveraging on this equivalence between the robust control and and the other seven. [05:04:34] Again, to start really from the 80s is nothing particularly new. And then it comes out if you saw these things that this shares looks like a logistic. Okay, and then you can put it much more easily into the code. [05:04:46] And then as I said before, because now into into risk, I was putting not only the sector productivity and the firm productivity, but I can put also the exchange rate or I can put also tariff uncertainty, which I can calibrate on the data. [05:05:02] One thing that I show, I mean, I'm not going to talk much about the intermediate step. The only thing that I want to show is that of course when you put capital, because those firms have an insurable risk. [05:05:13] Capital is going to decrease in proportion of the uninsurable risk. I mean, that's a quite obvious results when firms face an insurable risk. They have to they tend to when risk increases, they tend to invest more in cash, more in liquid assets. [05:05:29] And in fact, much of the increasing cash is typically interpreted as an increase in uncertainty and much less in capital. Okay, that is an independent result. [05:05:38] Okay, now to go to the full general equilibrium model. I have to do two things. [05:05:47] Or to the to the quantitative model, I have to match this gradual decline. And I couldn't do it for many months. But then I use the Julia matching and basically I put the kind of calvo not on the prices, but on the shares. [05:06:30] So the firms can reoptimize those shares only at infrequent time. And this allows me to go spot on on the matching with the importer responses with the households and investors. Okay. [05:06:50] Now, Mark, this comes probably this will make you happier. Okay, so I had to put the investors inside. And so first thing I thought, well, I put the consumers, I mean, there's to be a dynamic optimization with the consumers. [05:07:02] And I put the classical hunk. So I put the beauty shock coming from the sky. But first of all, I never liked this beauty shock coming from the sky. And second of all, there is already shocks in the model, which is the productivity shocks of the firms. [05:07:15] Okay, and so there is a very nice way to tie them, or at least one that I like. That is, you can use this range sharing contract coming from the limited commitment contract. [05:07:26] Okay, so the risk is coming from the firm. In my way, there is now plenty of evidence that actually the beauty shocks is basically firm risk shock. Okay. So the firm is choosing the wage and the firm is going to ensure the worker. [05:07:41] Okay, but only up to a certain point. The firm will choose wages. This is a classic implicit contract type of wages to maximize profits subject to three type of constraints. [05:07:53] The first one is that the expected utility of the worker is at a certain level. Okay, this is the insurance part. And then there are two constraints that are state contingents. [05:08:02] Now the key one is, I mean, this is the workers and this is the firm. Now the key one is the firm. Why? Because when the firm binds, the firm cannot ensure the workers anymore. [05:08:13] And so the extent of an insurability within the general equilibrium model is given by the complete market kernel glass, the large multiplier on this constraint, because this tells me the regions. [05:08:28] I mean, if the shock is large enough, the firm cannot cope with the shock and says, okay, guys, I can not ensure your wage. Now you're going to do this sharing with me and you're going to be hit by the shock. [05:08:40] And so how the extent to which this binds determines how much risk is going to be transferred to the workers. Now the workers are also investors. So then the kernel will be their own marginal productivity. [05:08:53] But as you know, the marginal, sorry, the MPC and the stochastic discount factor of the workers, which can be written as an MPC as I will show you in a moment. [05:09:05] Okay, so I mean, what is very important is that so the wage will be flat when the constraint of the firm is not binding. And when it binds, the workers will receive some of the firm's shock. Okay. [05:09:17] And the way I can, it's calibrated of course to hit certain parameters, the only risk insurance parameters that I could find in the literature is the one that we so Pista Ferry and Schibarty have estimated on the imps data. [05:09:30] So I use that, but I think it's pretty sound. Okay, so then the kernel becomes a stochastic discount factor of every worker. You can log linearize this with the same log linear algebra is going to depend on the on the region in which the firm constraint is binding and it's going to depend on the MPC of the workers. [05:09:47] Okay, there's a bunch of things floating around. [05:09:50] Now how do I this small open economy? How do I close this small open economy with the rest sharing condition I mean the classical buckles and Smith and then I have net foreign asset position. Okay, I'm not doing gross foreign asset position. [05:10:02] So I'm not going to have the reevaluation from the precision but I have net foreign asset position. [05:13:21] Again, it depends on where I put the shock. So if I put the shock, just the tariff and the volatility, but I put it only on the covariance. So this covariance that pops up in the misallocation of firms. [05:13:34] I don't go down very much. But once I put the uncertainty shock also on the on the investors kernel, then I managed to go down much more. How do I get the gradual thing through the Julia, how are you see and [05:13:50] how do I get the share of firms that do not change their relationship with I introduced, which I introduced as a calva. The shares I mean much of it goes home. [05:14:02] They declined both the China one and the rest of the world, the real exchange rate appreciate I mean we know that there's been some depreciation afterwards but this was also due to other shocks, net foreign assets don't move much. [05:14:14] The second thing which I think would be interesting is that I compute welfare and I mean of course because you are the heterogeneous agents, I can look at the distributional consequences of those shocks. [05:14:25] Okay, so typically in trade they have a lot of interest for distributional consequences of the China shop automation shops and so on and so forth. [05:14:33] In trade they never have all those stochastic things. [05:14:37] Okay, so those are good welfare numbers. Okay, so I can divide workers depending because they get the risk through the limited commitment contract. Okay, and by the way, and the model is calibrated so that there are eight sectors and three origins and the [05:14:54] eight sectors are calibrated on the shares from from OECD. So there are some workers that are working in the low China share, China firms that are having low median exposure to China and employers that work in the high in firms with high exposure to China. [05:15:14] And as you can see, those ones are the ones that are hit the most. [05:15:19] And I mean if you go into really the ones that are working in those sectors and the high China shares, the loss is pretty significant. Let me say that in this kind of scenario, I'm assuming that all the revenues from the tariff are wasted. [05:15:35] Okay, so they are put into some government spending but the government spending doesn't give a new utility. But if I give back the revenues to the workers, actually there might be some gains, but the distribution stays the same. [05:15:48] So all the workers that are working in the exposed sectors, they get a loss. [05:15:53] Okay, so to conclude. [05:15:56] What's having this food on the computer? ## Discussion (05:17:00 – 05:30:00) [05:17:52] Well, there's a lot going on, but the first thing is the high level of questions. [05:21:17] I have a question. [05:21:41] People have papers here that I'm not putting on the list, [05:21:44] but I just wanted to sort of give you some ideas [05:21:46] of what's out there. [05:21:48] So market specifics on costs of exporting, [05:21:51] experienced dependent costs or demand, [05:21:56] cost of the accumulation of customers or search, [05:21:58] or buyer networks which are chosen in advance. [05:22:01] So this is an example of an input-output paper [05:22:04] where aspects of the buyer network are chosen in advance. [05:22:08] So that's kind of the exporter side. [05:22:10] You could also do it on the importer side. [05:22:12] So models with inventories, [05:22:15] models with some costs of importing instead of exporting. [05:22:19] This Eton et al paper shows up twice [05:22:22] because the search frictions, [05:22:24] it's a two-sided as a matching problem. [05:22:26] So you can think about that, [05:22:27] shows up on both the buyer and the seller [05:22:29] are subject to this matching friction papers with, [05:22:32] I think here I'm maybe missing Chevinum's work, [05:22:37] input networks chosen in advance of shocks. [05:22:41] There are a series of papers, for example, [05:22:43] this fan and low paper, this Bolivar paper, [05:22:47] they're papers which are very close to your work [05:22:49] where they involve choice of the input-output network [05:22:52] ex ante under uncertainty within complete markets. [05:22:56] So those are papers that are sort of very relevant [05:23:00] to this paper. [05:23:17] Or kind of again, [05:23:27] so that's what's on the literature. [05:23:32] I don't think that's going to happen. [05:24:53] And it's kind of a bit of a, [05:25:40] so I'm just wondering what's the equivalent [05:25:43] of this issue, that's kind of long. [05:25:46] Obviously, there are ones that don't matter, [05:25:51] but it is all for you, it's continuous. [05:25:53] They are going to go to the next one. [05:25:55] So obviously, it's easy. [05:26:00] easy. [05:26:04] Thank you for the time. [05:26:06] It's no good to talk about the path of the circumstances. [05:26:09] It's not good to talk about the situations that we are experiencing. [05:26:11] I'll be the next speaker. [05:26:13] Thank you. [05:26:20] Please. [05:26:22] Thank you. [05:26:24] Thank you. [05:26:26] Thank you. [05:26:28] Thank you. [05:26:30] Thank you. [05:26:32] Thank you. [05:26:36] Thank you. [05:26:38] So in this picture, [05:26:40] I think that's amazing. [05:26:42] Either, [05:26:44] how much does that return [05:26:46] to their listing? [05:27:17] I think it's a lot of money [05:27:24] by the way. [05:27:59] I don't know. [05:28:01] I just want [05:28:03] to get a comment by this particular case. [05:28:05] In our standard assumption is that firms use the margin [05:28:12] utility [05:28:14] and the utility utility utility, [05:28:16] which is a very important point, [05:28:18] is that firms use the margin utility of the representative household [05:28:22] to price different states. This kind of risk aversion or kind of curvature that's [05:28:28] heterogeneous across firms is somehow less standard. I wanted to point out that you can [05:28:34] still have a role for incomplete insurance across countries. So that's kind of the nice thing [05:28:39] about the country setting is that, you know, if you don't have complete markets across [05:28:44] countries, you can still have a role for incomplete insurance to play to be important for deciding [05:28:51] the network under this endogenous network here. So you don't necessarily need to put it in at the [05:28:55] firm level. And then I think Esther did mention what she has in the data, the disciplines. This [05:29:02] is this Guiso et al paper, what disciplines the heterogeneous risk pricing across firms [05:29:06] within the country. It would be nice if there was something that was more tightly mapped [05:29:10] to the actual sort of risk price, something that more closely captures, you know, risk pricing of firms [05:29:20] than something that's coming out of sort of a labor paper to discipline this. And I have a [05:29:25] question as to whether it's important for quantitative results to go to the firm heterogeneity and [05:29:30] whether, you know, something like in this Bolivar paper would be enough. So I think that's [05:29:37] basically all I have to say. I do have one last slide, but it disappeared.